The amount of recorded seismic event data is rapidly growing, and manual processing by trained human experts to infer hypocenter, source parameters, and moment tensor solutions is therefore no longer feasible. Automated procedures are required to process data efficiently and include quality‐control measures that allow for outlier detection. We present a modular cross‐correlation location (CCLoc) algorithm for induced seismicity that uses cross correlations of either raw seismograms or characteristic functions derived from them followed by a reverse migration procedure. The novelty of this approach is the inclusion of cross pairs of P and S arrivals and the inclusion of autocorrelations, both of which add a distance constraint to the hypocenter estimation. The algorithm is modular in the sense that preprocessing can be tailored to specific data or task.
Nine months of seismic data from an underground hard‐rock tin mine are processed in a fully automated mode using a machine‐learning approach for seismic phase arrival detection and using the estimated arrival functions as input for CCLoc. Making use of the average cross‐correlation value as a quality constraint, CCLoc can successfully infer source information on 92% of previously manually processed data. The accuracy of automatic processing is demonstrated by comparing hypocenter, source parameter, and moment tensor solutions between the two datasets. The algorithm will potentially aid the analysis of induced or other seismicity and is particularly well suited to use in the case of large numbers of seismic sensors recording many events.